Moderation Analysis using Multiple Gr... PreviousNext
Mplus Discussion > Structural Equation Modeling >
 GradStudent posted on Monday, March 14, 2011 - 4:49 pm
I need help to figure out the syntax for the following:

Independent Variable is Treatment
I want to find out if the change in depression from pre to posttreatment is moderated by sex and several other dychotomous variables.

I tried using the syntax from some other users that have posted here, but I am getting an error.

my basic regression is

Depression Post on Depression Pre

but I don't know how to constrain the path.

i need to get the unconstrained paths for males and females

and then the constrained paths for males and females

and calculate the chi square diff test.

I would really appreciate any help with this. Ideally, if there is a syntax already written on this, I could then replace the variables to match mine.

Thank you!
 Linda K. Muthen posted on Tuesday, March 15, 2011 - 10:59 am
You could do a multiple group analysis with gender as the grouping variable. The unconstrained model is:

post ON pre;

The constrained model is:

post ON pre (1);
 Jodie Stearns posted on Wednesday, October 07, 2015 - 11:36 am
I am running a mediation model (1 IV, 2 mediators, 2 DVs) and also testing for moderation by gender. It is a complex design with ordinal mediators so I am using WLSMV. I have 2 questions:
1) To test for moderation my plan is to use a multiple group model and (a) set all pathways in the male and female group to be equal to test whether the whole model is equal across gender, and (b) set the pathways to be equal one by one to test whether each pathway is equal across gender. Is this the procedure you would recommend?
2) When I compare the multiple group model (H1) to a constrained model with all pathways set as equal (H0) using DIFFTEST I find that there is a significant difference in model fit. Typically I would interpret this as a worsening of model fit and retain the H1 model however I notice that there appears to be an improvement in model fit based on the fit statistics (see below). Should I interpret this as a significant improvement in fit?
H1: Chi-square = 1865.245, df = 2, p < .000; RMSEA = .360, p < .000; CFI = .360
H0: Chi-square = 1752.020, df = 17; RMSEA = .122, p < .000; CFI = .404
Diff test = 106.286, df = 15, p < .000
 Bengt O. Muthen posted on Wednesday, October 07, 2015 - 5:39 pm
1) I would set them equal and let Modindices tell me which ones aren't.

2) You should go by DIFFTEST and not compare the individual fit values. But note that both H0 and H1 fits very poorly - in this case chi-square difference testing such as DIFFTEST isn't reliable.
Back to top
Add Your Message Here
Username: Posting Information:
This is a private posting area. Only registered users and moderators may post messages here.
Options: Enable HTML code in message
Automatically activate URLs in message